Available online at www.sciencedirect.com ScienceDirect

Geochimica et Cosmochimica Acta 148 (2015) 130–144 www.elsevier.com/locate/gca

Direct measurement of production rates by (a,n) reactions in minerals

a, a b Stephen E. Cox ⇑, Kenneth A. Farley , Daniele J. Cherniak

a Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, CA 91125, USA b Department of Earth and Environmental Sciences, Rensselaer Polytechnic Institute, Troy, NY 12180, USA

Received 2 April 2014; accepted in revised form 27 August 2014; available online 21 September 2014

Abstract

The production of nucleogenic neon from capture by 18O and 19F offers a potential chronometer sensitive to temperatures higher than the more widely used (U–Th)/He chronometer. The accuracy depends on the cross sections and the calculated stopping power for alpha particles in the mineral being studied. Published 18O(a,n)21Ne production rates are in poor agreement and were calculated from contradictory cross sections, and therefore demand experimental verification. Similarly, the stopping powers for alpha particles are calculated from SRIM (Stopping Range of Ions in Matter software) based on a limited experimental dataset. To address these issues we used a particle accelerator to implant alpha particles at precisely known energies 21 18 into slabs of synthetic quartz (SiO2) and tungstate (BaWO4) to measure Ne production from capture by O. Within experimental uncertainties the observed 21Ne production rates compare favorably to our predictions using published cross sec- tions and stopping powers, indicating that ages calculated using these quantities are accurate at the 3% level. In addition, we 22 21  measured the Ne/ Ne ratio and (U–Th)/He and (U–Th)/Ne ages of Durango fluorapatite, which is an important model sys- tem for this work because it contains both and fluorine. Finally, we present 21Ne/4He production rate ratios for a variety of minerals of geochemical interest along with software for calculating neon production rates and (U–Th)/Ne ages. Ó 2014 Elsevier Ltd. All rights reserved.

1. INTRODUCTION produced by these reactions is also readily apparent in crus- tal well gases (Kennedy et al., 1990). Yatsevich and Honda Alpha particles produced by the decay of and (1997), who were primarily interested in the neon budget of series interact with light elements in min- Earth’s mantle and atmosphere, proposed the use of nucle- erals to produce the three stable of neon. Nuclides ogenic neon as a chronometer. They established that the created by such secondary nuclear reactions are termed reaction 18O(a,n)21Ne dominates the nucleogenic produc- nucleogenic, and can be produced by alpha particle capture tion of neon in crust and mantle, and that the reaction and also by capture of emitted by alpha particle 24Mg(n,a)21Ne is insignificant in comparison. In addition, capture. Wetherill (1954) was the first to recognize produc- using published capture cross section measure- tion of nucleogenic neon, from (a,n), (a,p), and (n,a) reac- ments they arrived at a mean bulk crustal production rate 21 4 8 tions in minerals containing uranium and thorium. Neon ratio Ne/ He = 4.5 10À , which is consistent with cur- Â rent estimates and measurements for U- and Th-bearing minerals, yet is much higher than earlier estimates (Kyser and Rison, 1982). Using similar data, Leya and Wieler ⇑ Corresponding author at: Caltech, MC 100-23, Pasadena, CA 21 22 91125, USA. Tel.: +1 (626) 395 2936. (1999) calculated the production rate of Ne and Ne in E-mail addresses: [email protected] (S.E. Cox), farley@gps. specific minerals of geochemical interest, such as apatite caltech.edu (K.A. Farley), [email protected] (D.J. Cherniak). and zircon. Gautheron et al. (2006) modeled the production http://dx.doi.org/10.1016/j.gca.2014.08.036 0016-7037/Ó 2014 Elsevier Ltd. All rights reserved. S.E. Cox et al. / Geochimica et Cosmochimica Acta 148 (2015) 130–144 131 rate of 18O(a,n)21Ne on the basis of a different analysis of (Yatsevich and Honda, 1997; Leya and Wieler, 1999; the published cross section data and arrived at production Gautheron et al., 2006). However, a known rate of neon rates that differed by up to 10% from those of Leya and production is critical for any chronometric application,  Wieler (1999). They also presented (U–Th)/Ne age determi- and existing estimates based on neutron production mea- nations on several apatite and zircon samples using these surements are not in sufficient agreement for high-precision production rates and considered the consequences for dat- geochronology (Leya and Wieler, 1999; Gautheron et al., ing of the spatial separation of parent nuclides from both 2006). In this paper we take a different approach than pre- and neon daughter products due to alpha ejection vious workers to establishing 21Ne production rates. We and the energy-dependence of the neon production rate. measure the neon itself in target phases exposed to energetic The (U–Th)/Ne geochronometer is potentially useful in alpha particles of precisely known energy. This work is rapidly cooled samples and as a chronometer of the mini- analogous to the oxygen gas target experiments of mum age of mineral formation. In the former case, one sim- Hu¨nemohr (1989) but undertaken on solid phases. We also ply assumes that the rock cooled so quickly that diffusive loss present new high precision U–Th–He–Ne data on Durango of neon has been negligible for its entire history, and in the apatite, which confirm the validity of our 21Ne results in a latter case, one acknowledges that diffusive loss may have natural system and also provide a production rate estimate occurred and accounts for this possibility in the geologic for the 22Ne reaction from alpha capture by 19F. interpretation. For example, Farley and Flowers (2012) For (U–Th)/Ne chronometry, we are primarily inter- applied (U–Th)/Ne dates as a minimum age for the forma- ested in the two reactions 18O(a,n)21Ne and 19F(a,n)22- tion of a hematite specimen from the Grand Canyon. Na(b+)22Ne (simplified as 19F(a,n)22Ne hereafter due to As discussed by Gautheron et al. (2006), (U–Th)/Ne the insignificance of the 2.6 year 22Na decay over geologic ages can also be used for thermochronometry. Ideally a time). These reactions occur at a rate proportional to the thorough understanding of the temperature dependence of number of alpha particles emitted and to the concentration neon diffusion supports such an application. However, of the target (18O or 19F) in the mineral. Further- because laboratory diffusivity measurements are usually more, the rate of these reactions also depends on the cross made under environmental conditions (temperature, pres- section of the target nuclide for alpha particles and the sure, chemical activity) much different from the natural set- chemical composition of the host mineral. Below, we ting, care is required in interpreting the laboratory data. describe the physical controls on this reaction and how they This is especially true for some phases of interest for are measured or calculated in order to arrive at a nucleo- (U–Th)/Ne dating, such as apatite, because phase modifica- genic neon production rate. tion via volatile loss in a high temperature vacuum chamber Neon production from (n,a) reactions on Mg is possible (Nadeau et al., 1999). As an alternative approach, can be safely ignored for minerals that do not contain large approximate closure temperatures for neon can be obtained amounts (wt%) of Mg. For example, O is about three times by comparison of neon ages to those of other thermochro- as productive of neon as Mg, so these production pathways nometers in the same rock (Gautheron et al., 2006). only become significant at the 1% level for Mg/O > 0.03 (in the case of 21Ne). is even less productive of 1.1. The production of neon by uranium and thorium decay neon than F, so these production pathways become signif- icant at the 1% level for Mg/F > 0.07 (in the case of 22Ne). In the (U–Th)/He system, multiple parents (uranium- In minerals that include magnesium as a stoichiometric 238, uranium-235, and thorium-232) decay through chains constituent, these reactions must be considered in the neon in which they emit a series of alpha particles with discrete budget. energies. -147 is also an alpha emitter, but The rate of a given depends on the because it produces just one alpha particle and because of nuclear properties of the target nuclide and the energy of its very low energy, this nuclide contributes little to nucleo- the incident particle, and is characterized by the nuclear genic production. The vast majority of alpha particles come cross section parameter (r(E)). The cross section is an areal to rest as 4He atoms, but a small number (fewer than one in representation of the probability of reaction in a pure sam- ten million) react with other elements in a mineral to form ple of the target nuclide and is frequently described in units 24 2 nuclear reaction products. A particularly important reac- of millibarns (1 barn = 10À cm ). A larger cross section tion is 18O(a,n)21Ne, which produces measurable neon in means that the reaction is proportionally more likely to many oxide and silicate minerals (Yatsevich and Honda, occur when a single particle interacts with the target, or 1997; Leya and Wieler, 1999; Gautheron et al., 2006; proportionally more frequent when a large number of par- Farley and Flowers, 2012). The reaction 19F(a,n)22Ne is ticles interact with the target. Over the energy range of even more productive per alpha particle, but is important interest here, cross sections for neon-producing alpha parti- only in minerals that contain significant fluorine, such as cle reactions generally increase with alpha energy. Cross apatite, titanite, or fluorite (Sole´ and Pi, 2006). Although sections for neon producing reactions have been measured only limited data exist, all analyzed minerals are more over the last several decades (Bair and Willard, 1962; retentive of neon than of helium (Shuster and Farley, Hansen et al., 1967; Bair and Del Campo, 1979; Norman 2005; Tournour and Shelby, 2008a,b; Behrens, 2010; et al., 1984), usually using a thin-target approach in which Cherniak et al., 2014). This suggests the possibility of using the cross section is determined by measuring the production measured neon concentrations in minerals as a chronome- rate of neutrons using alpha particles at just a single energy. ter like (U–Th)/He, but retentive to higher temperatures In contrast, nucleogenic neon production in a mineral 132 S.E. Cox et al. / Geochimica et Cosmochimica Acta 148 (2015) 130–144

3 involves reactions occurring along the entire trajectory as Finally, we multiply by 10À to convert mbarn to barn, 24 2 3 the alpha particle slows and eventually comes to rest, and 10À to convert barn to cm , and 10À again to convert mg thus involves an integral along the alpha particle energy to g, which allows us to arrive at the following complete path. These observations have three important implica- yield equation (e.g., Yatsevich and Honda) tions: neon production will be higher for alpha particle Eak 30 NA N18 r E À decays with high characteristic energy; neon production will Y 21Ne 10 X0 ð Þ dE 1 ¼ AZ N0 SZ E ð Þ depend on how fast alpha particles lose energy within a Z0 ð Þ given mineral; and neon production will preferentially This equation has units of atoms and reflects the pro- occur early in the alpha particle trajectory, where energies duction rate of neon for a single alpha particle k of initial energy Eak. This quantity will be quite small, about are highest. 8 5 10À , because most alpha particles come to rest as Stopping power S(E) describes the rate of energy loss of 4 Â a given particle as it travels through a given material. Stop- He without reaction. ping power depends on the interactions of incident particles As a concrete example, we illustrate this calculation for quartz (SiO2) and barium tungstate (BaWO4) in Fig. 1. with both the nuclei and electron clouds of the atoms in the 18 21 host mineral. In practice the electronic stopping power is Fig. 1a shows the cross section for the reaction O(a,n) Ne several orders of magnitude larger than the nuclear stop- as a function of alpha particle energy, with original data ping power and dominates at energies of interest here. Rel- points in black circles and the smoothed interpolation pre- ativistic effects can be ignored here because alpha particles sented as a red line (Section 1.2), Fig. 1b shows the stopping produced by are well below the energy at power for alpha particles in quartz as a function of alpha which such effects become significant (about 35 MeV). particle energy (Ziegler et al., 2010), Fig. 1c shows the inte- r E ð Þ Stopping power is represented with density-normalized grand S E as a function of alpha particle energy, and Fig. 1d Zð Þ units of MeV/(mg/cm2) to avoid the requirement of mea- shows the full yield equation given stoichiometric 18 18 suring density in natural samples. Like the cross section, XO = 0.5326 and d OVSMOW = +2 ( O/O = 0.002005). stopping power is dependent on the energy of the incident The yield equation for 21Ne from alpha particles in alpha particles, so it must also be considered as an integral quartz (Fig. 1d) demonstrates several notable features. quantity as these particles lose energy and come to rest. The production rate increases toward the higher energy Many studies have measured and modeled stopping powers end of the alpha energies produced by uranium and tho- in great detail (e.g., SRIM and references therein; Ziegler rium decay, so these decays will dominate the total nucleo- et al., 2010). genic neon in a sample. Conversely, alpha emitters with no Reaction cross section, oxygen-18 or fluorine-19 con- high-energy decays (147Sm, for example) will produce negli- tent, and stopping power of a given mineral are the basic gible nucleogenic neon. In addition, because alpha particles ingredients required to predict the neon production rate are more productive when first emitted (at their highest from alpha-emitting . Production is linearly energy), most nucleogenic neon will be produced closer to proportional to the concentration of the target nuclide in the alpha emitter than the final rest position of 4He that the host mineral, and is proportional to an energy-depen- does not react. This means that the alpha ejection correc- dent integral quantity that includes both the cross section tion for neon produced by alpha particles will be smaller and the total stopping power for alpha particles in the host than the alpha ejection correction for radiogenic helium mineral. We will take this integral from rest (Ea = 0 MeV) in the same minerals (Gautheron et al., 2006). to the energy of each alpha particle emitted during . As an example, in the next section we construct 1.2. Physical quantities used to predict neon production rates the yield equation for 21Ne from 18O for a given alpha par- ticle in a generic mineral Z. For use in chronometry it is both impractical and unnec- 18 We represent the concentration of O in Z as the prod- essary to measure neon production rates at many alpha uct of the mass fraction of oxygen (XO) in the mineral and energies for a large number of minerals. One may instead 18 the isotopic abundance of O (N18/NO), divided by the use a (validated) cross section and stopping powers calcu- total of one formula unit of the mineral lated from SRIM, along with the stoichiometric oxygen (AZ). We multiply this quantity by the Avogadro constant concentrations and assumed or measured oxygen isotopic (NA) to arrive at (XO)(N18/NO)(NA/AZ), which has units of compositions of any mineral to calculate a neon production atoms/gram. rate. In this section we predict 21Ne production rates in syn- ‘The integral quantity that describes the rate of reaction thetic materials using these methods, and then compare the as an alpha particle interacts with the target mineral scales prediction to our measured production rates in order to val- with the cross section (a higher cross section means more idate the cross sections and stopping powers. We chose the production) and scales inversely with stopping power (a minerals quartz (SiO2) and barium tungstate (BaWO4) higher rate of energy loss means less neon production). because we were able to obtain large, optically pure syn- As a consequence our integrand is the cross section r(E) thetic crystals of each and because the stopping power for divided by the mineral-dependent stopping power SZ(E). alpha particles in quartz is almost twice as high as in bar- We then integrate this from rest (Ea = 0 MeV) to the energy ium tungstate, so we can easily deconvolve the effects of of the incident alpha particle Eak. We arrive at the term, cross section and stopping power on the measured neon Eak r E 2 ð Þ dE which has units of (mbarn)(mg/cm ). production rates. 0 SZ E ð Þ R S.E. Cox et al. / Geochimica et Cosmochimica Acta 148 (2015) 130–144 133

Fig. 1. An illustration of the steps involved in calculating 21Ne yield from alpha particles in quartz (SiO2; black) and barium tungstate 18 21 (BaWO4; green). The figure shows (a) the interpolated cross section of the reaction O(a,n) Ne, including the renormalized (Bair and Del Campo, 1979) data of Bair and Willard (1962) and the high energy data of Hansen et al. (1967), (b) the density-normalized stopping power for 21 alpha particles in quartz and barium tungstate, (c) the integrand r(E)/SZ(E) from the yield equations, and (d) the calculated yield for Ne from alpha particles in quartz and barium tungstate. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

The concentration of 18O in a mineral exhibits linear Gautheron et al. (2006), for our predicted 21Ne yields we control on the production rate of neon in the reaction use a 0.01 MeV interpolation of the renormalized Bair 18O(a,n)21Ne, so it is crucial to account for stoichiometric and Willard (1962) data below 5.14 MeV because it is the variations in oxygen composition and to note the oxygen most detailed, and we supplement it with the more sparse concentration used in calculating reported (U–Th)/Ne ages. data from Hansen et al. (1967) in the energy range above In the case of measured differences in d18O or small varia- 5.14 MeV (Fig. 1a). tions in oxygen concentration, the production rates pro- We also adopt the cross section curve for the 19F(a,n)22Ne vided here may be scaled linearly because the changes in reaction presented by Murata et al. (2006), which is based alpha particle stopping powers will be small. The mass frac- on a small amount of very sparse experimental thin target tion of oxygen in mineral phases is usually known from data. As noted by Murata et al. (2006), numerous reso- stoichiometry, and the variations in oxygen isotopic com- nances were observed in this reaction below alpha energies position measured in natural minerals are usually small of 3.1 MeV by Wrean and Kavanagh (2000). Although compared to other uncertainties in the (U–Th)/Ne system. these resonances are too low energy to be important for We measured d18O = +25& for our synthetic barium tung- neon production, similar features must exist in the energy state and d18O=0& for our synthetic quartz using the range of interest (primarily 3–8 MeV, because alpha parti- laser fluorination method at Caltech (Eiler et al., 2000). cles with higher energy are not emitted by uranium and tho- 18 We assume a common igneous d OVSMOW = +6& rium series decay, and alpha particles with lower energy (18O/O = 0.002012) for the other calculations presented produce very little neon; Fig. 1) for (U–Th)/Ne dating as here unless otherwise noted. The user may modify this well, but have not been documented. Thus the 19F(a,n)22Ne value in the software provided in the appendix. cross section should be used with more caution than the The 18O(a,n)21Ne cross section used by (Gautheron cross section for the reaction 18O(a,n)21Ne owing to the et al., 2006) is based on a critically evaluated compilation sparseness of these measurements. of data from several sources and includes numerous We computed the stopping powers of the minerals of resolved resonances. The thin target experiments vary in interest using SRIM (Ziegler et al., 2010) with stoichiome- quality and include a dataset from Bair and Willard tries found on webmineral.com. The primary limitations (1962) that was later renormalized by a factor of 1.35 by of SRIM are that it depends on experimental data that (Bair and Del Campo, 1979) because the original thin can vary in quality and that it does not account for crystal target thicknesses were measured incorrectly. Following structure but rather assumes an amorphous phase. 134 S.E. Cox et al. / Geochimica et Cosmochimica Acta 148 (2015) 130–144

However, experimental data for alpha particles are rela- 19F is the only isotope in natural fluorine, and the cross sec- tively plentiful, and the stopping of ions is dominated by tion of neon production from 19F is higher than for 18O. electronic collisions that will not be greatly affected by crys- Using Durango fluorapatite as a model system, we explored tal structure. Effects such as channeling in a crystal lattice the potential use of 22Ne produced by 19F as an indepen- can be neglected due to very low probability and low dent chronometer and as a counterpart to 21Ne produced impact on total neon production. While we suspect errors by 18O in the same phase. Even with a less well-known pro- introduced to our calculations from the stopping powers duction rate, 22Ne is potentially useful as a relative chro- from SRIM are small, our experimental measurements, nometer (Sole´ and Pi, 2006) or as a target for diffusivity involving phases with very different predicted stopping studies in fluorine-rich minerals, and is useful for deducing powers, allow us to assess this expectation. the presence of a fluorine-bearing phase (e.g., an inclusion) The calculated neon production rates for other minerals in a studied sample. allow users to calculate (U–Th)/Ne ages from uranium, thorium and neon data for a wide variety of minerals of 1.5. Samarium-147 and cosmogenic neon geochemical interest (Section 4.3). As explained above, oxy- gen concentration and stopping power both affect the pro- We do not consider the impact of 147Sm or of cosmo- duction rate of nucleogenic neon in a given mineral, so the genic neon production in the calculations in this paper. differences between minerals include both of these effects. These are irrelevant for our synthetic samples, in which we control the sole production pathway for neon, but they 1.3. Neon-21 production rates measured by alpha particle might be relevant in Durango fluorapatite and other natu- implantation ral minerals that contain significant rare earth elements and/or that have been exposed to cosmic radiation. How- As mentioned above, previously published 21Ne produc- ever, 147Sm emits a single alpha particle with energy of only tion rates estimated for uranium- and thorium-containing 2.23 MeV. The neon production rate at this energy is negli- minerals differ by up to 10% (Leya and Wieler, 1999; gible due to the extremely low cross section for the reac-  18 21 19 22 Gautheron et al., 2006), which is greater than the uncer- tions O(a,n) Ne and F(a,n) Ne in this energy range tainty of most measurements. This discrepancy (Fig. 1; Wrean and Kavanagh, 2000). is sufficiently large as to preclude accurate chronometry Cosmogenic neon dating has applications to many sur- and is primarily related to different interpretations of sparse face processes, especially to ancient surfaces that cannot and contradictory 18O(a,n)21Ne cross section data (Bair be dated as effectively using radioactive nuclides such as and Willard, 1962; Hansen et al., 1967; Bair and Del 10Be. Durango fluorapatite is mined from below the surface Campo, 1979). Note that our calculations in Section 1.1 so should be free of this component, but cosmogenic neon use the same approach for establishing the 18O(a,n)21Ne is a potential contaminant for samples that have been cross section as Gautheron et al. (2006), so are not an inde- exposed within 1 m of Earth’s surface for a substantial  pendent estimate for comparison. of time, so it should be considered during sampling To assess the accuracy of the predicted 21Ne production and in cases for which the rock history is unknown. Produc- rates, we directly measured the rates by implanting a known tion rates of cosmogenic neon are mineral-dependent, but fluence of alpha particles of a single energy into thick tar- are typically of the order of a few tens of atoms per gram gets (i.e., mineral samples in which all alpha particles stop) per year. Cosmogenic neon is produced with a distinctive of synthetic materials and then measuring the amount of isotope ratio (mineral-dependent, but roughly subequal in 21Ne produced. all three isotopes; Niedermann, 2002) compared to air While quartz and barium tungstate are not likely to be and nucleogenic neon, so one can usually recognize it by targets of (U–Th)/Ne chronometry, they were appealing analyzing all three neon isotopes. When helium is also ana- choices for several reasons. As noted above, their extreme lyzed, the presence of 3He can be an indicator of cosmo- contrast in stopping power may allow us to distinguish genic nuclide contamination. errors in stopping power from errors in cross section. In addition, we were able to obtain large optical grade crystals 2. EXPERIMENTAL in both cases, which ensured homogenous samples of ade- quate size for the experiment. We experimentally verified 2.1. Sample selection and preparation that the samples were free of neon and helium before per- forming the implantations. Measured quantities are exclu- We prepared 1 cm square and 1 mm-thick slabs of   sively a product of the implantation experiment. quartz and barium tungstate by cutting with a wire saw and cleaning in an ultrasonic bath of 18.2 MX purified 1.4. A natural test of the 21Ne production rate and water to ensure that no residue from the cutting slurry preliminary 22Ne production rate from Durango apatite was present. These samples are far thicker than the average stopping distance for alpha particles in either mineral ( 15–  is a challenging target for (U–Th)/Ne chro- 30 lm depending on energy) and therefore represent thick nometry as it substitutes readily for hydroxyl and targets that effectively stop all incident alpha particles. in many minerals and is therefore often not stoichiometric. The quartz and barium tungstate analyzed in this study In addition, it is more difficult to measure accurately than had low concentrations of non-nucleogenic (air-like) neon, oxygen, e.g., by electron probe. However, the target nuclide with nucleogenic neon typically representing >85% of the S.E. Cox et al. / Geochimica et Cosmochimica Acta 148 (2015) 130–144 135 measured 21Ne. In minerals that tend to release more air monitor the beam, and assess whether there was any contam- neon, such as hematite (Farley and Flowers, 2012), the ination of the incident helium beam by . This could uncertainty is often limited by the correction for air con- be evaluated by observation of changes in the backscatter sig- tamination, which effectively raises the background level nal if the spectra included features characteristic of against which the nucleogenic neon signal must be resolved. hydrogen backscattered from the elemental constituents of Finally, the presence of non-atmospheric neon from mantle the sample, along with spectra produced by the backscat- or crustal fluid sources could complicate nucleogenic neon tered helium. measurements, and should be considered in the case of mea- sured isotope ratios that do not fall on an air-nucleogenic 2.3. Neon mixing line. We prepared the Durango fluorapatite samples from 2.3.1. Neon mass spectrometry crushed and homogenized aliquots of a single gem-quality We measured neon using a GV Helix SFT static vacuum crystal such as the ones used for (U-Th)/He development magnetic sector noble gas mass spectrometer with a Balzers (Farley, 2000). The crushing and homogenization proce- SEV-217 ion counting multiplier and a mass resolution (R) dure ensures that the uranium, thorium, neon, and helium of approximately 800 (m/Dm, Dm = peak width at 5% of distributions are homogenous between samples even maximum peak height). To extract neon from the samples, though the large crystal may have uranium and thorium we first degassed them in a double vacuum furnace at tem- zonation as reported by Boyce and Hodges (2005). peratures of 1100–1300 °C, and then removed other gases using non-evaporable getters, a liquid cryogenic 2.2. Alpha particle implantation trap, and a focusing cryostat. We isolated neon on a cryo- stat at 21 K and then pumped away helium to avoid detri- We implanted alpha particles using the 4 MV Dynami- mental effects of many orders of magnitude more helium tron linear accelerator at SUNY-Albany. In order to pro- than neon in the mass spectrometer. Some helium is duce alpha particles with energies in the range of retained at this nominal temperature on our cryostat, but approximately 4–8 MeV, which is the range encountered we observe that the majority is removed. Neon-21 beams in the decay chains of 238U, 235U, and 232Th, we used dou- were several hundred counts per second (cps), well above bly charged particles accelerated at 2–4 MV. Using the cal- multiplier backgrounds of 0.15 cps and procedural blanks  culations described in Section 1, we chose the alpha particle of 3 cps. We measured each neon isotope beam on the  fluence for each energy by calculating the fluence necessary multiplier by peak hopping with the accelerating voltage to achieve an easily measurable quantity of neon. The alpha at a fixed magnetic field to avoid hysteresis effects and set- particle fluence we used ranged from 1014 particles for the tling delays. Between analyses, we measured air standards  15 highest energy (most productive) experiments to 10 par- in the same way as the samples to keep track of mass dis-  ticles for the lowest energy experiments. crimination and sensitivity in the mass spectrometer. We We identified the helium beam produced by the Dynam- use isotopic deconvolution and the measured composition itron using the spectra of backscattered particles from a of air standards to remove air backgrounds from samples. synthetic silica slab, which produce a characteristic spec- We are able to pseudo-resolve doubly charged 40Ar from trum of backscattered alpha particles in measurements with 20Ne (R = 1777 for 40Ar++ and 20Ne+, 45% resolved, iso-  a solid-state surface barrier detector. After focusing the bar contribution less than 1& at measurement position on beam, we placed the synthetic target (either quartz or bar- 20Ne) adequately enough to measure 20Ne without signifi- ium tungstate) in the path of the incident beam and then cant interference when concentrations are controlled measured accumulated charge in order to achieve the with the liquid nitrogen cryogenic trap. Doubly-charged desired alpha particle fluence. Spectra of backscattered par- dioxide overlaps almost completely with 22Ne ticles from each sample during the implantations were also (R = 6229), but we reduce its small contribution signifi- monitored and collected with a solid-state surface barrier cantly by lowering the electron energy in the ionization detector to ensure that only an alpha particle beam was source for 22Ne measurement only, which reduces double implanted into samples during the experiments. We also ionization of carbon dioxide far more than the single ioni- monitored the shape and position of the incident beam dur- zation of neon. Carbon dioxide backgrounds are stable and ing each experiment to ensure complete collection of beam less than 5& of 22Ne signals, and we remove any carbon on the targets. The Dynamitron has an energy resolution dioxide with the two steps of cryogenic separation. 4 (DE/E) of about 10À . The nominal uncertainty of this mea- surement is quite low (better than 1 ppm), but the practical 2.3.2. Neon-21 measurement by isotope dilution and sample- uncertainty is limited to about 2% by the purity of the standard bracketing helium beam (Section 2.4). We made 21Ne measurements by isotope dilution using a At energies above 7 MeV, we experienced problems with spike gas artificially enriched in 22Ne relative to air and to beam stability and associated inadvertent implantation of the samples to ensure that pumping away the helium does hydrogen along with the helium beam. This hydrogen does not affect the neon concentration determinations. Our not produce neon, but is included in the integrated beam cur- apparatus and spike gas for this application changed over rent measurement along with the helium. During the implan- the course of the experiments described here. tation, a solid-state surface barrier detector set at 167.5 The measured isotopic composition of our first spike gas degrees with respect to the incident beam was used to was approximately 8.9% 20Ne, 0.12% 21Ne, and 91% 22Ne. 136 S.E. Cox et al. / Geochimica et Cosmochimica Acta 148 (2015) 130–144

The improved spike was of >99.5% pure 22Ne. This gas is 2.4. Helium introduced into the vacuum line just after completion of the neon extraction in the furnace. After spiking, the mea- For approximately half of the implanted samples we sured mixture comprises an extraction line background measured for neon, we used the integrated implantation including isobaric interferences, air contamination in the current as the measure of the alpha particle fluence. After samples and vacuum system, the sample gas, and the spike it became apparent that two high-energy samples were gas. We measured the spike composition separately and, unusually poor in neon compared to the expected produc- after subtracting the extraction line background from all tion rate, we began to suspect that hydrogen had been other measurements, completely resolve the three compo- implanted with helium at the highest energies. In addition, nents by matrix inversion using the three isotope measure- in the quartz target at 7.8 MeV, two replicate runs showed ments. The sample concentrations were determined by evidence of significant hydrogen contamination in the RBS comparing spiked sample isotope ratios to spiked mano- spectra obtained during implantation (see Section 2.2). As a metrically calibrated air standards. consequence, we then switched to a method that allowed us The presence of measurable 20Ne and 21Ne in the first to split a very small amount of the sample helium and then spike added to experimental error on both air corrections measure it on a separate mass spectrometer. A split was and 21Ne determinations. In addition, we discovered a required because the full sample of implanted helium is sev- problem with the hardware (a virtual leak in the pipette vol- eral orders of magnitude larger than we can measure. Mea- ume) that led to spurious amounts of spike being emitted suring helium directly allowed us to determine correct during a small number of experiments. Fortunately, this neon/helium ratios even in the samples for which the alpha problem was easily detected in 22Ne measurements, and particle beam instability of the Dynamitron resulted in a could be corrected by using the 22Ne excess to determine significant implantation of hydrogen along with the alpha how much extra spike we added. However, this method particles. relies on consistent sample-standard measurements and We isolated 0.18% of each helium sample in a small vol- therefore defeats the purpose of using isotope dilution. ume after gas extraction, then analyzed this gas on an MAP For this reason, we simply subtract spike and air contami- 215-50 static vacuum magnetic sector noble gas mass spec- nation 21Ne based on the 22Ne and 20Ne measurements for trometer connected to the same extraction system as the these few samples and present results using conventional GV Helix SFT used to measure neon. This split represents sample-standard bracketing to determine absolute abun- an insignificant amount of the neon sample but, due to the 8 21 dance. For the vast majority of samples, sample-standard 5 10À production rate ratio of Ne to alpha particles,  Â bracketing and isotope dilution yielded statistically indistin- contained enough helium to achieve a signal of tens to hun- guishable results, which confirms that our procedure for dreds of millivolts on a Faraday cup detector equipped with pumping away helium does not remove or fractionate neon. an amplifier with a 1011 ohm resistor. We analyzed gas stan- In our final results, we report isotope dilution results except dards using the same split volume, so that uncertainties in when spiking problems suggest the use of the sample-stan- the volume ratio are removed from the calculated helium dard bracketing results. concentrations. We performed all helium analyses by sam- We made the Durango fluorapatite measurements using ple-standard bracketing with a helium-doped air standard both sample-standard bracketing and isotope dilution after developed at Caltech (“Caltech air”). Blank corrections we redesigned our isotope dilution hardware and spike gas, for helium were insignificant. which allows us to introduce a manometrically calibrated The uncertainties on helium measurements are about aliquot of the >99.5% pure 22Ne that is repeatable to better 1%, and are dominated by gas standard calibration. We than 0.1%. We made the first set of measurements without assigned a 2% uncertainty to helium concentrations calcu- the addition of spike gas to characterize the isotopic com- lated from the integrated beam current on the Dynamitron position of the neon from the fluorapatite, then we made rather than measured directly. The nominal uncertainty of the second set of measurements using the same method as the implantation is far smaller (of order ppm), so this value used for the implantation measurements, but we compared is somewhat arbitrary and is meant to account for the typ- the sample concentrations directly to the calibrated spike ical observed scatter between the measured helium concen- instead of using sample-standard bracketing. trations and the integrated beam currents in samples that The analytical uncertainties of neon measurements are show no evidence of hydrogen contamination. We assume about 1.5% (2r) with isotope dilution, dominated by that this added uncertainty is dominated by small amounts counting statistics and gas standard calibration, and of contamination of the beam with hydrogen, and by mea- typically about 2% without isotope dilution. For the two surement uncertainty on the mass spectrometer. samples that required a significant correction to the sam- ple-standard bracketing result due to the problem with 3. RESULTS the isotope dilution apparatus, the uncertainty is about 4%. The higher uncertainty of the latter is primarily due 3.1. Neon and helium in synthetic targets to the correction for the added spike, which we calculated using its isotopic composition. We increased the stated Neon and helium measurements and 21Ne/4He ratios in uncertainties of these samples to encompass the results the synthetic quartz and barium tungstate targets are shown from both methods, so some data points have asymmetric in Table 1. Nucleogenic 21Ne/4He ratios are displayed in uncertainties. Fig. 2 for quartz and Fig. 3 for barium tungstate. S.E. Cox et al. / Geochimica et Cosmochimica Acta 148 (2015) 130–144 137

Table 1 Neon and helium data for implanted quartz and barium tungstate. Column “Ne Method” indicates whether we used isotope dilution for the neon measurement (“ID”) or rejected it due to a hardware problem in favor of sample-standard bracketing (“No ID”). Column “He Method” indicates whether we calculated the amount of implanted helium from the integrated beam current on the Dynamitron with 2% assumed uncertainty (“Current”) or from a direct measurement by mass spectrometry (“MS”). Last two columns give the + and uncertainties À separately since they are asymmetric. All uncertainties are 1 standard deviation. Alpha Ne He 21Ne 21Ne ± 4He 4He ± 21Ne/4He 21Ne/4He 21Ne/4He 8 Energy method method (Mat) (Mat) (Tat) (Tat) (10À ) + À Quartz 4.5 No ID Current 16.5 1.8 1163 23 1.58 0.08 0.17 5.00 No ID Current 13.3 1.5 684 14 2.63 0.13 0.68 5.25 ID Current 14.3 0.6 467 9 3.06 0.12 0.12 5.50 ID Current 14.0 0.6 397 8 3.51 0.14 0.14 6.00 ID Current 18.2 0.7 343 7 5.30 0.21 0.21 6.25 ID Current 18.5 0.8 300 6 6.17 0.25 0.25 6.50 ID Current 18.4 0.8 264 5 6.96 0.28 0.28 7.00 ID Current 18.9 0.8 210 4 9.00 0.36 0.36 7.30 ID Current 16.5 0.7 170 3 9.72 0.39 0.39 7.80 ID Current 10.4 1.1 140 3 7.47 0.82 0.82 7.80 ID Current 14.8 0.2 140 3 10.6 0.15 0.15

BaWO4 4.5 ID MS 9.09 0.15 892 2 1.02 0.02 0.05 5 ID MS 12.2 0.2 710 1 1.72 0.03 0.06 5.25 ID MS 10.9 0.2 570 1 1.92 0.03 0.10 5.5 No ID MS 8.36 0.13 481.7 1.0 2.43 0.12 0.69 6 ID MS 11.3 0.2 335.2 0.7 3.38 0.05 0.05 6 ID MS 10.6 0.2 332.7 0.7 3.18 0.05 0.05 6.25 ID MS 12.1 0.2 291.8 0.6 4.13 0.07 0.07 6.5 ID MS 12.9 0.2 293.8 0.6 4.40 0.07 0.07 7 ID Current 13.4 0.2 250 5 5.35 0.08 0.08 7.3 ID MS 11.5 0.2 184.4 0.4 6.22 0.10 0.57 7.8 ID MS 16.5 0.3 209.1 0.4 7.89 0.47 0.12

Fig. 2. Comparison of measured 21Ne yields per alpha particle in quartz with predictions based on the published cross sections of Gautheron et al. (2006) and stopping powers described in Section 1.2. The measured values at 7.8 MeV are internally inconsistent and probably represent uncorrected contamination of the alpha particle beam with hydrogen. Also shown are the relative distributions of alpha particle energies in each of the uranium and thorium decay chains. 138 S.E. Cox et al. / Geochimica et Cosmochimica Acta 148 (2015) 130–144

The 21Ne concentrations are between 8 and 19 million on splits of the same crushed sample of Durango fluorapa- atoms in each sample. For the samples in which it was tite following the procedures outlined in House and Farley directly measured, 4He concentrations range from (2000), and Table 5 shows (U–Th)/He age determinations 140 1012 to 1163 1012 atoms. With two exceptions, on a subset of samples measured with a larger helium split   these numbers agreed with the total measured beam cur- for increased accuracy. Because of the significant correction rent. As expected, in both targets the 21Ne/4He ratios for nucleogenic 22Ne in fluorine-rich Durango fluorapatite, increase monotonically with alpha particle energy. The only we observed no improvement in precision by switching to exceptions are the two data points at Ea = 7.8 MeV in isotope dilution. quartz (Fig. 2); these are the samples with a large amount of contamination of the beam with hydrogen, prompting 4. DISCUSSION the use of the helium measurement method. The total span 8 7 in production rate ratio is 1.58 10À to 1.06 10À in 4.1. Comparison with predicted production rates 8  8  quartz and 1.02 10À to 7.89 10À in barium tungstate.   At a given energy the production rates are higher in quartz We observed good agreement (typically < 3% difference) than barium tungstate. All of these observations are in between measured 21Ne/4He production rate ratios and the accord with predictions outlined in Section 1.2. values predicted from the cross section data compiled in Gautheron et al. (2006) for quartz and barium tungstate. 3.2. Neon, helium, uranium, and thorium in Durango In contrast, we observed poor agreement when using fluorapatite Leya and Wieler’s (1999) values (up to 10% difference).  The agreement for both analyzed phases indicates that the Three sets of experiments were done on splits of the stopping powers and 18O(a,n)21Ne cross sections are both Durango fluorapatite sample. In the first and third sets accurate at the level of better than 3%. If the stopping pow- helium and neon were measured by peak height comparison ers were incorrect, one or both phases would exhibit 21Ne while in the second set helium was measured by peak height production rates inconsistent with predicted values, and comparison but neon was measured by isotope dilution the difference would not be systematic. If the cross section (Tables 2 and 3). We used the measured 21Ne/22Ne ratio were incorrect, we would observe a systematic difference in the unspiked data set for the isotope dilution calculations between measured and predicted production rates in each (Table 2). These samples used our improved isotope dilu- target regardless of stopping power. tion apparatus. All neon measurements are shown together The quartz data differ by an average of 2.8% from the in Fig. 4. Finally, Table 4 shows uranium and thorium mea- modeled production rates, with the modeled rates slightly surements made by solution ICP-MS using isotope dilution higher. The barium tungstate data differ by 0.6% with the

Fig. 3. Comparison of measured 21Ne yields per alpha particle in barium tungstate compared with predictions based on the published cross sections of Gautheron et al. (2006) and stopping powers described in Section 1.2. Also shown are the relative distributions of alpha particle energies in each of the uranium and thorium decay chains. S.E. Cox et al. / Geochimica et Cosmochimica Acta 148 (2015) 130–144 139

Table 2 Neon isotope data of Durango fluorapatite. SE – standard error of the mean. Uncertainties on individual analyses reflect 1r analytical error, while population uncertainties reflect 2r standard error. When isotope dilution is used, 21Ne concentrations calculated with isotope dilution (ID) and without are shown for comparison. Number Mass (mg) 21Ne 21Ne ± 21Ne ± (%) 22Ne 22Ne ± 22Ne ± (%) 22Ne/21Ne 22Ne/21Ne ± 22Ne/21Ne ± (%) (Mat/g) (Mat/g) (Mat/g) (Mat/g) 1 77.32 242 5 2.0 5165 45 0.87 21.3 0.6 2.6 2 75.04 239 5 1.9 5153 44 0.85 21.6 0.6 2.6 3 73.51 244 5 1.9 5228 44 0.85 21.4 0.5 2.5 4 87.05 249 5 1.9 5321 44 0.83 21.4 0.5 2.5 5 84.85 247 5 1.9 5303 44 0.83 21.5 0.5 2.5 6 88.77 244 5 1.8 5307 43 0.81 21.8 0.5 2.4 7 92.58 245 5 1.9 5296 43 0.82 21.7 0.5 2.5 8 110.9 247 5 1.9 5328 44 0.82 21.6 0.5 2.4 Average 1–8 244 5262 21.5 SE (2r) 0.88% 0.95% 0.46% Number Mass (mg) 21Ne (Mat/g) 21Ne ± (Mat/g) 21Ne ± (%) 21Ne (Mat/g) ID 21Ne ± (Mat/g) ID 21Ne ± (%) ID 9 97.27 244 4 1.8 241 6 2.3 10 78.66 246 5 1.8 246 6 2.4 11 108.95 243 4 1.8 240 5 2.3 12 95.51 249 4 1.8 246 6 2.3 13 86.36 243 4 1.8 245 6 2.3 14 91.82 248 4 1.8 248 6 2.3 15 109.39 245 4 1.8 243 6 2.3 16 108.22 244 4 1.8 243 6 2.3 Average 9–16 245.25 244.01 SE (2r) 0.63% 0.80% Number Mass 21Ne 21Ne ± 21Ne ± 22Ne 22Ne ± 22Ne ± 22Ne/21Ne 22Ne/21Ne 22Ne/21Ne (mg) (Mat/g) (Mat/g) (%) (Mat/g) (Mat/g) (%) ± ± (%) 17 90.0 255 1 1.6 5646 6 0.3 22.13 0.12 1.63 18 203 259 1 1.1 5793 4 0.2 22.33 0.08 1.09 19 26.1 242 2 3.0 5286 12 0.7 21.84 0.22 3.01 20 806 262 1 0.5 5732 2 0.1 21.86 0.04 0.55 21 28.2 238 2 2.9 5252 11 0.6 22.05 0.21 2.91 22 161 251 1 1.2 5711 5 0.3 22.77 0.09 1.24 23 237 258 1 1.0 5742 4 0.2 22.30 0.08 1.01 24 139 252 6 6.7 5507 21 1.1 21.89 0.50 6.79 25 479 250 4 5.2 5554 16 0.9 22.24 0.39 5.27 Average 17–25 252 5580 22.16 SE (2r) 2.1% 2.4% 0.9%

modeled rates slightly lower. These relationships may the accepted age, and serves primarily as a check on the imply that the cross section is correct and that the quartz procedures used for noble gas and uranium and thorium stopping powers are 3.0% too low, but the standard devi- analyses. ation of the measured to modeled production rate ratios The three sets of Durango fluorapatite experiments is about 4% for both targets, so the difference for both is reported here varied in methodology. The first two sets within the uncertainty of the measurements. each contained eight samples of roughly the same mass, with the only difference being the use of isotope dilution 4.2. Durango fluorapatite systematics for the second set of samples. The third set contained six samples of broadly varying mass chosen to explore the We find that the (U–Th)/21Ne age of the Durango fluor- source of the unexpectedly high (U–Th)/21Ne ages of apatite is 34.5 ± 3.3 Ma. This is in agreement with the the first two sets. The third set of analyses was made accepted age of the standard (31.44 ± 0.18 Ma; McDowell without isotope dilution, a necessary first step in analyz- et al., 2005), albeit with an uncertainty larger than we ing fluorapatite samples because of the presence of natu- believe to be obtainable with the (U–Th)/21Ne system, for ral 22Ne. In the third set, it is clear that the ratio of reasons outlined below. For several of the third set of Dur- measured neon signal to neon concentration in the mass ango fluorapatite measurements, we simultaneously mea- spectrometer varies with sample size (Fig. 4), and that sured (U–Th)/He ages of the same aliquots and obtained only the smallest samples provide ages close to the a mean age of 32.6 ± 0.7 Ma (Table 5). This value is near accepted age of the Durango fluorapatite standard. The 140 S.E. Cox et al. / Geochimica et Cosmochimica Acta 148 (2015) 130–144

Table 3 Helium and Ne/He data for Durango fluorapatite. Uncertainties on individual analyses reflect 1r analytical error, while population uncertainties reflect 2r standard error. 4 4 4 21 4 8 21 4 8 22 4 8 22 4 8 Number Mass (mg) He He ± He ± (%) Ne/ He (10À ) Ne/ He (10À )± Ne/ He (10À ) Ne/ He (10À )± (Tat/g) (Tat/g) 1 77.32 5499 55 1 4.40 0.10 93.9 1.3 2 75.04 5518 55 1 4.33 0.09 93.4 0.9 3 73.51 5550 56 1 4.40 0.09 94.2 0.9 4 87.05 5578 56 1 4.46 0.10 95.4 0.9 5 84.85 5519 55 1 4.47 0.10 96.1 0.9 6 88.77 5566 56 1 4.38 0.09 95.3 0.9 7 92.58 5539 55 1 4.41 0.09 95.6 0.9 8 110.9 5561 56 1 4.44 0.09 95.8 0.9 Average 1–8 5541 4.41 95.0 SE (2r) 0.35% 1% 1% Number Mass 4He 4He ± 4He ± 21Ne/4He 21Ne/4He 21Ne (ID)/4He 21Ne (ID)/4He 8 8 8 8 (mg) (Tat/g) (Tat/g) (%) (10À ) (10À )± (10À ) (10À )± 9 97.3 5550 56 1 4.39 0.09 4.34 0.11 10 78.7 5512 55 1 4.47 0.09 4.46 0.11 11 109.0 5494 55 1 4.42 0.09 4.37 0.11 12 95.5 5583 56 1 4.46 0.09 4.41 0.11 13 86.4 5503 55 1 4.42 0.09 4.44 0.11 14 91.8 5536 55 1 4.47 0.09 4.48 0.11 15 109.4 5563 56 1 4.40 0.09 4.36 0.11 16 108 5515 55 1 4.43 0.09 4.41 0.11 Average 9–16 5532 4.43 4.41 SE (2r) 0.40% 0.52% 0.81%

Fig. 4. Durango fluorapatite 21Ne (left axis) and 22Ne (right axis) concentrations per gram of sample for all measured aliquots, plotted against sample size. cause of this correlation is not clear, and is under inves- tion, helium is present at a concentration approximately tigation, but it may be related to ion source effects from eight orders of magnitude higher than that of neon), or incomplete removal of helium (prior to cryogenic separa- to other gases extracted from fluorapatite at high S.E. Cox et al. / Geochimica et Cosmochimica Acta 148 (2015) 130–144 141

Table 4 Uranium, thorium, and (U–Th)/Ne age data for Durango fluorapatite based on multiple aliquots for uranium and thorium and the population averages for neon for the first two datasets. Uncertainties on individual analyses reflect 1r analytical error, while population uncertainties reflect 2r standard error. Sample Mass (lg) U (ppm) U ± (ppm) Th (ppm) Th ± (ppm) Th/U Ne Age (Ma) Ne Age ± (Ma) A 133.6 9.21 0.12 173.9 1.2 18.8 33.8 1.4 B 257.3 8.99 0.13 169.8 1.1 18.9 34.6 1.4 C 298.2 8.87 0.15 170.2 1.2 19.2 34.7 1.4 D 100.9 8.79 0.13 167.4 1.0 19.0 35.3 1.4 Average A-D 8.97 0.18 170 2.7 19.0 34.6 2.4 E 27.37 9.51 0.11 183.7 0.6 19.3 32.5 1.3 F 29.54 9.33 0.10 173.3 0.6 18.6 35.1 1.4 G 39.33 9.20 0.06 175.2 0.6 19.0 35.2 1.4 H 35.57 9.20 0.08 176.0 0.6 19.1 35.1 1.4 I 35.12 9.38 0.08 178.1 0.6 19.0 34.0 1.4 Average E-I 9.32 0.13 177.3 4.0 19.0 34.4 2.1

Table 5 Helium and (U–Th)/He age data for part of the third dataset, using population averages of uranium and thorium. For these analyses, we increased the split size of the helium in order to obtain accurate (U–Th)/He ages. Number Mass (mg) 4He (Tat/g) 4He ± (Tat/g) 4He ± (%) He Age (Ma) He Age ± (Ma) 5 28.23 5441 54 1 32.62 0.49 6 161.18 5435 54 1 32.58 0.49 7 236.62 5500 55 1 32.98 0.50 temperature. We note that for samples without nucleo- the code. This software has not been evaluated by Geochi- genic 22Ne, isotope dilution offers a path around such mica et Cosmochimica Acta or its reviewers. problems. In Table 6 we provide mineral-specific production rates Apatite (U–Th)/22Ne chronometry depends on the con- of 21Ne and 22Ne per alpha particle for the decay chains centration of fluorine in the apatite, which is much more starting with 238U, 235U, and 232Th assuming secular equi- variable than the oxygen content. However, the production librium. These production rates can be used to calculate rate of 22Ne is much higher than that of 21Ne because while (U–Th)/Ne ages or to interpret neon/helium ratios in these 19 22 8 the cross section for F(a,n) Ne is about half that of minerals. Production rates average around 4.5 10À per 18 21 19 21 Â O(a,n) Ne, the isotope abundance of F (100%) is much alpha particle for Ne for all three decay chains in each higher than that of 18O( 0.2%). For example, we measure mineral. Using this value and assuming a Th/U ratio of 22 21  a Ne/ Ne production rate ratio of 21.52 ± 0.14 in Dur- 1, we estimate the age and uranium and thorium concentra- ango fluorapatite despite the fact that the F/O ratio is tions required to obtain measurable 21Ne in a mineral approximately 0.09. The exact relationship between F/O (Fig. 5). In order to plot this result in two dimensions, we and 22Ne/21Ne production will vary slightly with oxygen use “equivalent uranium” (alpha activity represented as isotope composition, sample age, and Th/U ratio, but will concentration of uranium; Gastil et al., 1967) and define be approximately (22Ne/21Ne)/(F/O) = 0.0042 in near-end- “measurable neon” as 65,000 atoms, or about 1.5 cps on member fluorapatite. our mass spectrometer. This requirement assumes a very In addition to its use as a chronometer, (U–Th)/22Ne low amount of non-nucleogenic 21Ne; if a large air correc- may be combined with (U–Th)/21Ne in the same mineral tion is required for an analysis, more nucleogenic neon is as a method for determining the fluorine content. If isotope required to achieve good precision (Section 2.3.2). dilution with 22Ne is not used, 22Ne can also be measured to investigate contamination with apatite or another fluorine- 4.3.1. Zircon, xenotime, thorite, and coffinite containing phase (e.g., titanite or fluorite). The isostructural tetragonal minerals zircon and xeno- time are commonly used for (U–Th)/Pb and, in the case 4.3. Neon production rates in other uranium- and thorium- of zircon, (U–Th)/He geochronometry and thermochro- bearing minerals based on validated cross section nometry and may also be of interest for (U–h)/Ne chro- nometry. Based on comparison with ages of other The appendix includes MATLAB code that can be used thermochronometers, Gautheron et al. (2006) estimated a to compute nucleogenic neon production rates and (U–Th)/ neon closure temperature in zircon of 400 ± 50 °C. Closure Ne ages for a variety of minerals. Users may define the ura- temperatures for the (U–Th)/Ne system intermediate nium, thorium, helium, and neon contents and d18O of min- between the helium and systems in these minerals raise erals. The user can also change the cross section, stopping the possibility of adding it to multiple chronometer thermo- power, mineral stoichiometries, and oxygen contents within chronology studies, and the (U–Th)/Ne system may also be 142 S.E. Cox et al. / Geochimica et Cosmochimica Acta 148 (2015) 130–144

Table 6 Average 21Ne and 22Ne production rates calculated based on verified production rates in a variety of minerals for each decay chain for stoichiometric compositions (obtained at webmineral.com) and d18O = +6&. Production rates assume secular equilibrium for each decay chain. Mineral Target Composition Production rates for each decay chain from Production rates for each decay chain from 18O* 19F* [O] [F] 238U 235U 232Th 238U 235U 232Th 21Ne/4He 21Ne/4He 21Ne/4He 22Ne/4He 22Ne/4He 22Ne/4He 8 8 8 8 8 8 (wt (wt (10À ) (10À ) (10À ) (10À ) (10À ) (10À ) fraction) fraction) Quartz 0.5326 0 4.04 5.62 6.08 – – – Barium 0.1662 0 2.44 3.39 3.66 – – – Tungstate Zircon 0.3491 0 3.38 4.70 5.08 – – – Xenotime-(Y) 0.3480 0 3.34 4.65 5.03 – – – Thorite 0.1974 0 – – – Coffinite 0.1953 0 2.85 3.95 4.26 – – – Titanite 0.4080 0 3.33 4.63 5.01 – – – Baddeleyite 0.2597 0 2.90 4.04 4.36 – – – Hematite 0.3006 0 2.80 3.89 4.21 – – – Maghemite 0.3006 0 2.80 3.89 4.21 – – – Goethite 0.3602 0 3.14 4.37 4.73 – – – Magnetite 0.2764 0 2.60 3.62 3.91 – – – Calcite 0.4795 0 3.59 5.00 5.41 – – – Aragonite 0.4795 0 3.59 5.00 5.41 – – – Uraninite 0.1185 0 2.17 3.01 3.24 – – – Thorianite 0.1212 0 – – – Cryptomelane 0.3485 0 3.14 4.37 4.72 – – – Fluorapatite 0.3801 0.0352 3.02 4.20 4.54 42.1 64.5 73.9 Fluorite 0 0.4867 – – – 595 915 1044

Fig. 5. Relationship between age and effective uranium concentration (eU = U + 0.235 Th, accounting for alpha dose over time) required 21 Â 21 4 8 for a measurable amount of Ne (defined as 65,000 atoms) in a generic mineral with a fixed Ne/ He production rate ratio of 4.5 10À , Â assuming a Th/U ratio of 1. The different lines reflect different sample sizes. The actual production rate will vary slightly according to mineralogy and Th/U ratio. S.E. Cox et al. / Geochimica et Cosmochimica Acta 148 (2015) 130–144 143 used as a tool to expose or investigate problems with these high common lead concentrations prevent the use of the chronometers in low-temperature and volcanic systems. (U–Th)/Pb system on these minerals. Neon appears to Thorite and coffinite are less common, especially as euhe- be retained at surface temperatures in oxides, and dral crystals suitable for chronological studies, but may the (U–Th)/Ne system is therefore already in use as a geo- occur as inclusions or associated accessories in other miner- chronometer in hematite (Farley and Flowers, 2012). als of chronological interest, and may be of interest them- Because of the low uranium and thorium concentrations selves in unusual settings. in the minerals themselves, care must be taken to either exclude or account for inclusions of other minerals that 4.3.2. Titanite and baddeleyite contain these elements. Other common candidates for (U–Th)/Pb dating include titanite and baddeleyite, both of which may be targeted for 4.3.5. Calcite and aragonite (U–Th)/Ne dating for the same reasons as the silicates in Carbonates are similar to iron oxides in that low ura- Section 4.3.1. Replacement reactions between baddeleyite nium and thorium concentrations and common lead con- and zircon may be more likely to completely reset the tamination sometimes complicate the (U–Th)/Pb system (U–Th)/Ne system than the (U–Th)/Pb system, especially (Rasbury and Cole, 2009) while low and variable helium when the reaction occurs in the direction zircon to badde- retention complicates the (U–Th)/He system (Cros et al., leyite, because lead is very resistant to diffusive loss 2014). Neon production rates are insufficient for dating of (Cherniak and Watson, 2001). The different production most Pleistocene marine carbonates using the (U–Th)/Ne rates of neon in zircon and baddeleyite, arising from differ- system due to low uranium and thorium concentrations, ences in oxygen concentration, should also be noted but older deposits or deposits with unusually high uranium because they often occur in close association. and thorium concentrations are candidates for (U–Th)/Ne geochronology. 4.3.3. Uraninite, thorianite, and the implications of micro- inclusions 5. CONCLUSIONS Minerals such as uraninite and thorianite are unlikely to occur in large crystals with well behaved (U–Th)/Ne sys- We present direct measurements of the 21Ne production tematics, but may occur as micro-inclusions in other miner- rates of mono-energetic alpha particles in two targets, als of interest, especially as alteration products in which allows us to verify the 18O(a,n)21Ne cross section hydrothermal systems. Even very small inclusions will have and the stopping powers independently. These measure- a profound impact on the (U–Th)/Ne systematics of host ments agree well with the cross section data compiled by phases due to the very high uranium and thorium concen- Gautheron et al. (2006) and with stopping power calcula- trations of these minerals and the fact that neon production tions using SRIM, affirming the validity of the calculated peaks very close to the source of alpha particles. Further- production rates at the 3% level. These production rates  more, very uranium-rich phases will become radiation dam- are presented here along with software for the calculation aged and will probably exhibit anomalous diffusivities for of production rates in other minerals and of (U–Th)/Ne neon, as is observed in similar situations for helium ages. These results allow geochemists to apply the (U– (Hurley, 1954; Nasdala et al., 2004; Shuster et al., 2006; Th)/Ne system to questions about Earth history, including Shuster and Farley, 2009; Guenthner et al., 2013). medium-temperature thermochronometry and geochro- The phases above are typically very high in uranium nometry in rocks with high uranium or older than and/or thorium, and will therefore likely exhibit high nucle- 10 Ma, with a better understanding of the inherent uncer- 21  ogenic Ne concentrations relative to air contamination tainty in the system. and small amounts of cosmogenic contamination. The phases below typically have much lower uranium and tho- ACKNOWLEDGEMENTS rium concentrations, so air contamination in the vacuum line and air adsorbed to and trapped within the samples will We thank Don Burnett for extensive discussion about nuclear be a more significant factor affecting the precision of the chemistry and implantation techniques. We also thank Ce´cile Gau- theron and Samuel Niedermann for their thoughtful signed measurements, and samples must be more carefully selected reviews, as well as an anonymous reviewer for a third review, and screened for cosmogenic neon contamination. This and we thank Associate Editor Pete Burnard for his attentive applies especially to samples that form at or near the handling of the manuscript. This research was supported by the Earth’s surface, such as carbonates, where cosmogenic neon National Science Foundation through grant NSF-EAR-1144500 contamination is more likely. to KAF and through a Graduate Research Fellowship to SEC.

4.3.4. Hematite, goethite, and magnetite Iron oxides frequently contain ppm-level concentra- APPENDIX A. SUPPLEMENTARY DATA tions of uranium and thorium that are adequate for the production of radiogenic helium and neon (Lippolt The software used to model neon production rates and et al., 1993; Shuster et al., 2005; Farley and Flowers, to calculate neon ages is included and is also freely available 2012). These uranium and thorium concentrations and at www.gps.caltech.edu/~scox/neon_age_calc. 144 S.E. Cox et al. / Geochimica et Cosmochimica Acta 148 (2015) 130–144

Supplementary data associated with this article can be Leya I. and Wieler R. (1999) Nucleogenic production of Ne found, in the online version, at http://dx.doi.org/10.1016/ isotopes in Earth’s crust and upper mantle induced by alpha j.gca.2014.08.036. particles from the decay of U and Th. J. Geophys. Res. 104, 15439–15450. Lippolt H. J., Wernicke R. S. and Boschmann W. (1993) 4He REFERENCES diffusion in specular hematite. Phys. Chem. Minerals 20, 415– 418. Bair J. and Willard H. (1962) Level structure in Ne22 and Si30 from 18 21 26 29 McDowell F. W., McIntosh W. C. and Farley K. A. (2005) A the reactions O ( , n)Ne and Mg ( , n)Si . Phys. Rev. 128, 40 39 a a precise Ar– Ar reference age for the Durango apatite (U– 299–304. Th)/He and fission-track dating standard. Chem. Geol. 214, Bair J. K. and Del Campo J. G. (1979) Neutron yields from alpha- 249–263. particle bombardment. Nucl. Sci. Eng. 71, 18–28. Murata T., Matsunobu H. and Shibata K. (2006) Evaluation of the Behrens H. (2010) Noble Gas Diffusion in Silicate Glasses and (a, xn) reaction data for JENDL/AN-2005. Jpn. Atom. Energy Melts. Rev. Mineral. Geochem. 72, 227–267. Agency-Res., 78. Boyce J. and Hodges K. V. (2005) U and Th zoning in Cerro de Nadeau S. L., Epstein S. and Stolper E. (1999) Hydrogen and Mercado (Durango, Mexico) fluorapatite: insights regarding carbon abundances and isotopic ratios in apatite from alkaline the impact of recoil redistribution of radiogenic 4He on (U–Th)/ intrusive complexes, with a focus on carbonatites. Geochim. He thermochronology. Chem. Geol. 219, 261–274. Cosmochim. Acta 63, 1837–1851. Cherniak D. J. and Watson E. B. (2001) Pb diffusion in zircon. Nasdala L., Reiners P. W., Garver J. I., Kennedy A. K., Stern R. Chem. Geol. 172, 5–24. A., Balan E. and Wirth R. (2004) Incomplete retention of Cherniak D. J., Thomas J. B. and Watson E. B. (2014) Neon radiation damage in zircon from Sri Lanka. Am. Mineral. 89, diffusion in olivine and quartz. Chem. Geol. 371, 68–82. 219–231. Cros A., Gautheron C., Pagel M., Berthet P., Tassan-Got L., Niedermann S. (2002) Cosmic-ray-produced noble gases in terres- Douville E., Pinna-Jamme R. and Sarda P. (2014) 4He behavior trial rocks: dating tools for surface processes. Rev. Mineral. in calcite filling viewed by (U–Th)/He dating, 4He diffusion and Geochem. 47, 731–784. crystallographic studies. Geochim. Cosmochim. Acta 125, 414– Norman E., Chupp T., Lesko K., Grant P. and WoodruffG. (1984) 432. 22 19 Na production cross sections from the F(a, n) reaction. Eiler J. M., Crawford A., Elliott T., Farley K. A., Valley J. W. and Phys. Rev. C: Nucl. Phys. 30, 1339–1340. Stolper E. M. (2000) Oxygen isotope geochemistry of oceanic- Rasbury E. T. and Cole J. M. (2009) Directly dating geologic arc lavas. J. Petrol. 41, 229–256. events: U–Pb dating of carbonates. Rev. Geophys. 47, RG3001. Farley K. (2000) Helium diffusion from apatite: general behavior as Shuster D. L. and Farley K. A. (2005) Diffusion kinetics of proton- illustrated by durango fluorapatite. J. Geophys. Res. 105, 2903– 21 3 4 induced Ne, He, and He in quartz. Geochim. Cosmochim. 2914. Acta 69, 2349–2359. Farley K. A. and Flowers R. M. (2012) (U–Th)/Ne and multido- Shuster D. L. and Farley K. A. (2009) The influence of artificial main (U–Th)/He systematics of a hydrothermal hematite from radiation damage and thermal annealing on helium diffusion eastern Grand Canyon. Earth Planet. Sci. Lett. 359–360, 131– kinetics in apatite. Geochim. Cosmochim. Acta 73, 183–196. 140. Shuster D. L., Flowers R. M. and Farley K. A. (2006) The Gastil R. G., DeLisle M. and Morgan J. R. (1967) Some effects of influence of natural radiation damage on helium diffusion progressive metamorphism on zircons. Geol. Soc. Am. Bull. 78, kinetics in apatite. Earth Planet. Sci. Lett. 249, 148–161. 879. Shuster D. L., Vasconcelos P. M., Heim J. A. and Farley K. A. Gautheron C., Tassan-Got L. and Farley K. (2006) (U–Th)/Ne (2005) Weathering geochronology by (U–Th)/He dating of chronometry. Earth Planet. Sci. Lett. 243, 520–535. goethite. Geochim. Cosmochim. Acta 69, 659–673. Guenthner W. R., Reiners P. W., Ketcham R. A., Nasdala L. and 22 4 Sole´ J. and Pi T. (2006) Determination of the Ne / He ratio Giester G. (2013) Helium diffusion in natural zircon: radiation nucl rad in natural uranium-rich fluorite by mass. Phys. Rev. C: Nucl. damage, anisotropy, and the interpretation of zircon (U–Th)/ Phys. 74, 047601. He thermochronology. Am. J. Sci. 313, 145–198. Tournour C. C. and Shelby J. E. (2008a) Helium solubility in alkali Hansen L. F., Anderson J. D., McClure J. W., Pohl B. A., Stelts M. silicate glasses and melts. Phys. Chem. Glasses – Eur. J. Glass L., Wesolowski J. J. and Wong C. (1967) The ( , n) cross a Sci. Technol. Part B 49, 207–215. sections on 17O and 18O between 5 and 12.5 MeV. Nucl. Phys. Tournour C. C. and Shelby J. E. (2008b) Neon solubility in silicate A: Nucl. Hadr. Phys. 98, 25–32. glasses and melts. Phys. Chem. Glasses – Eur. J. Glass Sci. and House M. and Farley K. (2000) Helium chronometry of apatite Technol. Part B 49, 237–244. and titanite using Nd-YAG laser heating. Earth Planet. Sci. Wetherill G. (1954) Variations in the isotopic abundances of neon Lett. 183, 365–368. and argon extracted from radioactive minerals. Phys. Rev. 96, Hurley P. M. (1954) The helium age method and the distribution 679–683. and migration of helium in rocks. In Nuclear Geology (ed. H. Wrean P. and Kavanagh R. (2000) Total cross sections and Faul). Wiley and Sons. pp. 301–329. 19 22 22 22 reaction rates for F(a, n) Na, Ne(p, n) Na, and their Hunemohr H. (1989) Edelgase in U- und Th-reichen Mineralen ¨ inverses. Phys. Rev. C: Nucl. Phys. 62, 055805. und die Bestimmung der 21Ne-Dicktarget-Ausbeute der 18- Yatsevich I. and Honda M. (1997) Production of nucleogenic neon O(a n)21Ne- Kernreaktion im Bereich 4.0–8.8 MeV. Max- Á in the Earth from natural radioactive decay. J. Geophys. Res. Planck-Institut fur Chemie. 102, 10291. Kennedy B. M., Hiyagon H. and Reynolds J. H. (1990) Crustal Ziegler J. F., Ziegler M. D. and Biersack J. P. (2010) SRIM – The neon: a striking uniformity. Earth Planet. Sci. Lett. 98, 277– stopping and range of ions in matter (2010). Nucl. Instrum. 286. Methods Phys. Res., Sect. B 268, 1818–1823. Kyser T. K. and Rison W. (1982) Systematics of rare gas isotopes in basic lavas and ultramafic xenoliths. J. Geophys. Res. 87, 5611. Associate editor: Pete Burnard